#1 Introduction +
Where is the digital revolution?
Faculty of Humanities and Social Sciences
University of Lucerne
20 February 2025
What makes a device looking intelligent?
AI is a moving target with respect to …
An image segmentation by Facebook’s Detectron2(Wu et al. 2019)
Speech-to-Text 💬
Text-to-Speech 📣
Speech-to-Speech 🗣️
voice translation by SeamlessM4T v2 (Duquenne et al. 2023)
voice cloning by VALL-E (Wang et al. 2023)
Generate podcasts based on any text
Generate a songs following instructions
is a brand, large-language models (LLM) is the technology
generates fluent text, not necessarily truthful
is highly useful, although it understands little
what is tough for humans might be easy for the model; and vice-versa
is English-focused, multi-linguality is limited
generates non-reproducible outputs
generated text cannot be detected (except verbatim parts)
yesterday’s version might be different than today’s
LLMs are a tool, learn how to use it 👍
ChatBots challenge classic search engines
answer with source attribution instead of ranked snippet
blurring the line between search and generation
Agents pursuing more and complex tasks